Property-Based Design Methodology I:
Parameters Influencing Solubility
Robert DeWitte ACD/Labs,
Toronto, ON, Canada
Abstract
Many biological barriers between dose and exposure have been
shown to correlate with the logD (pH) of a compound. Such is certainly the case for aqueous solubility (see below), permeability, plasma protein binding, and tissue partitioning. To a degree, and for
certain classes of compounds, the same is true of blood brain barrier permeability, though in this case, it is generally observed that
polar surface area is the dominant molecular physical property.
Fedor Gorohov, Eduard Kolovanov ACD/Labs,
Moscow, Russian Federation
As a result, the increase in solubility as a function of pH is attenuated for these compounds. This observation provides a working hypothesis for identifying members of this structural class
with improved solubility: if the pKa can be lowered, so that a
greater extent of ionization can be achieved at pH 5.5, then the
solubility observed at that pH will be increased due to further
ionization.
Given these observations, one would like to ask the question:
how can I change my compound in order to improve the property of interest? Specifically, how can I decrease the logD value at
a specific pH?
This series of posters develops an approach for addressing these
questions using the drug acetyl sulfadiazine as an example. We
begin by reviewing the evidence supporting the claim that solubility correlates with logD, and use this to develop working
hypotheses for improving solubility. These hypotheses are then
translated into a general approach to improve the drug-like properties of medicinal compounds.
Solubility Correlates with
LogD: A Theoretical Argument
This dependence can be argued as follows: consider solubility as
a coupled pair of phenomena, where the first transition moves
from an ordered solid state to a disordered liquid state. One can
consider this intermediate state as either a super-cooled liquid, or
a solution of solute in itself. The second transition moves from
this ideal solution to a solution of solute in water. Therefore, the
first transition corresponds to isothermal melting, whereas the
second corresponds to partitioning. The magnitude of the
isothermal melting contribution to aqueous solubility will depend
on the architecture of the solid state, which is largely a property
of the sample, rather than the isolated chemical structure (e.g.,
crystalline vs. amorphous, polymorphs, salts, hydrates, etc. will
have a pronounced affect on this term). The magnitude of the
second transition will depend on the partitioning between ideal
solution and aqueous solution, and logD is a reasonable model
for this. [Note that logD does not correctly take into account the
influence of the solubilities of various salts, and so the solubility
of the fully ionized form in the GI could be quite different from
that modeled with logD]. Finally, the relative magnitude of these
terms can be gleaned by an evaluation of Yalkowsky’s model for
solubility, where logS~-logP-0.01mp. In short, an increase of 100
degrees in melting point is required to decrease solubility by one
order of magnitude, whereas this same degree of change can be
accomplished with a one log unit change in lipophilicity.
Solubility Correlates with
LogD: Experimental Evidence
Acetyl Sulfadiazine is a weak acid with pKa of 6.3, having an
intrinsic solubility of 146 µg/mL. Low solubilities are therefore
observed at relatively low pHs, such as 5.5. In this case, the compound is only 11% ionized. At higher pHs, however, the solubility increases greatly: at pH 7.4 the compound is 90% ionized and
the solubility is in the range of 2.5-3.6 mg/mL.
Because the pH of 5.5 is more relevant than 7.4 for the purposes
of achieving appreciable oral absorption, we will consider ways
to increase the solubility at 5.5 by making small changes to the
chemical structure of acetyl sulfadiazine.
Further experimental data for related compounds (methyl and
dimethyl substitution on the pyrimidine ring) provide further illustration of the influence of pKa on the pH profile of solubility.
In the case of 4-methyl substitution, the pKa has raised to 6.7
from 6.3. This small change is sufficient to alter the fraction ionized significantly—down from 11% to 5% at pH 5.5 and down
from 90% to 80% at pH 7.4. This effect is more extreme in the
case of the dimethyl analogue: a further half log unit increase in
pKa reduces the ionized fraction to 1% and 60% at pHs 5.5 and
7.4, respectively.
Turning our attention now to the intrinsic solubilities of analogues within this chemical series, we can see that the logarithm of the intrinsic solubility trends negatively with logP.
Recall that intrinsic solubility refers to the solubility of the neutral form of the compound. In terms of the pH profile of the solubility, the intrinsic solubility defines the plateau at the bottom
of the curve. (Note that zwitterions do not manifest their intrinsic solubilities because there is no pH at which the neutral form
is in 100% abundance).
Caveats
The above assertion that solubility correlates with logD, has its
limitations. Several mitigating factors can obfuscate this trend:
1. The upper limit of solubility differs from the lower limit of
logD. While the lower limit of logD(pH) is defined by ion-pair
partitioning, this is not the case for the upper limit of solubility.
a. Where there is no salt counter-ion present in solution,
the upper-bound on ion solubility is limited by the ability
of the solvent to support a very high ionic strength.
b. Where there is ample counter-ion in solution, the solubility is limited by the salt product formed with this solution.
2. The lower limit of solubility depends on more than logP.
Clearly, other factors, such as size, flexibility, symmetry, and
other supramolecular factors, contribute strongly to the
intrinsic solubility of a compound.
A General Procedure for
Designing Improved
Properties
In general, we could describe a suitable process in the following
algorithmic terms
1. Consider your SAR, and its relationship to molecular structure.
a. Ideally, there will be some region of the structure for
which the SAR is relatively flat with respect to substitution.
In such a case, attention will therefore be focused on how
to adapt these parts of the molecule to achieve the maximal improvement in physicochemical properties. This is
generally best accomplished through the addition or substitution of a small substituent.
b. In some cases, however, there is no such region of flat
SAR, and you are limited to working within the topology
of the parent compound. In this case, heterocycle substitution may be the only recourse available. This case is
taken up as the last example of this application where it is
shown that heterocycle replacement can be as effective
as the addition or substitution of a small substituent.
2. Predict the logD(pH) curve. Note the predicted values at
the key pH, and also the shape of the curve around that
point. One of three cases will be present:
a. The molecule contains a pKa within two log units of
the pH of interest, which, if moved closer, will reduce the
logD dramatically.
b. The molecule contains a pKa within two log units of the
pH of interest, which, if moved further will reduce the
logD dramatically.
c. The molecule does not have a pKa near the pH. In this
case, introducing an ionizable group with a suitable pKa
will reduce the logD dramatically.
3. For cases a & b, predict the pKa s of the parent molecule
and inspect the relationship between chemical structure and
the pKa to be moved. Look for a relationship between the
electron withdrawal strength of substituents and/or heterocycles and the acid/base strength of the ionization centre.
This is reported as a Hammett equation in the calculation
protocol window of ACD/pKa DB.
4. Note the sigma and Pi constants of any key substituents
and/or heterocycles influencing the pKa.
5. Consult the Substituents Databases within ACD/MedChem
Advisor to identify suitable replacements.
a. For Substituents, use “Neutral Substituents.CFD”
i. Query on a data range for the appropriate Sigma
Constants,
ii. Restrict molecular weight to below 40 or so,
iii. Restrict Pi values to be less than zero,
iv. Limit the view to substituents that have appeared in
at least one phase II compound.
b. For Heterocycles, use “Heterocycles.CFD”
i. Begin with a substructure search to limit the search to
heterocycles with similar topology to the one in the par-
ent compound,
ii. Query on a data range for the appropriate Sigma
Constants,
iii. Restrict Pi values to be less than zero,
iv. Limit the view to substituents that have appeared in at
least one phase II compound,
v. Compute the pKas of the heterocycle on the parent
compound.
• In the case that the heterocycle includes the pKa you are
seeking to modify, query also on ranges of either acidic or
basic pKa to identify suitable substitutes.
• In the case that the heterocycle does not include the pKa
you are seeking to modify, limit the ranges of acidic and
basic pKas of the heterocycles to those similar to the parent
heterocycle.
c. Consider synthetic feasibility, molecular stability, and
other factors
i. The features of ACD/ChemFolder will help you focus
and limit your browsing.
d. Select and assess key substituents. For each in turn,
i. Double-click in ChemFolder to bring the substituents
into ChemSketch,
ii. Attach the substituents or heterocycle to the parent
compound,
iii. Compute pKa, logD(pH),
iv. Assess logD at the pH of interest.
6. In the case where you seek to introduce an acidic or basic
centre, this can be achieved by consulting one of the following
databases:
a. “Heterocycles.CFD”
b. “Acidic Substituents.CFD”
c. “Basic Substituents.CFD”
d. Note that you can query all of them simultaneously
using the features of ACD/ChemFolder, and then work
within the result set as a separate database.
e. The above design principles can be applied similarly to
identify a substituent or heterocycle that introduces the
appropriate degree of ionization to your compound.
Note that it is always preferable to begin with experimental values
for the compound, or a related compound, and use this to verify
that the predictions from ACD/LogD Sol Suite are accurate, to within a reasonable confidence interval. In addition, it is a good idea to
inspect the calculation protocol for logP and pKa to determine if
the algorithm is using secondary approximations (logP coefficients
indicated as “Approximated”, “Estimated”, or “Assumed”), RingBreaking approximations (pKa DB), or extensive correction factors
in the pKa calculation protocol. In any of these situations, or if the
prediction error is not sufficiently low, it is recommended to invoke
one of the system training capabilities, whereby experimental
measurements are used to improve the calculation accuracy on
related compounds. This is described in detail in the Application
Note entitled “The Inner Workings of User Training.”
In general, given two predicted values (A and B) and their confidence intervals (±a and ±b), the difference between the two of
them (B-A) is only relevant if B-A is greater than the larger of
a or b. Using this rule of thumb, you can be certain that the predicted values supporting the design approach illustrated in this
application note will deliver recommendations that will be borne
out to within rank order. This means that the molecular changes
predicted to be improvements of the properties will indeed
improve the properties, though the extent of the improvement
may be greater or less than predicted.
Conclusions and Working
Hypotheses
Nonetheless, the observations presented here provide a framework for making decisions about structural analogues:
1. The effect of ionization on solubility is to modulate the
intrinsic solubility, providing a steep rise in solubility above the
intrinsic level as a function of pH. Therefore, in considering
structural changes within our chemical series, we need to
select substituents that increase the fraction of ionization
2. The effect of lipophilicity is to reduce solubility. Therefore,
the structural changes pursued in order to increase ionization
should be done in such a manner as to not significantly
increase the lipophilicity of the compounds.
90 Adelaide St. W., Suite 600
Toronto, Canada M5H 3V9
Tel: (416) 368-3435
Fax: (416) 368-5596
Toll Free: 1-800-304-3988
Email: [email protected]
w w w. a c d l a b s . c o m